FAYETTEVILLE, Ark. - A machine inspection system under development at the University of Arkansas (UA) Division of Agriculture can peer through intact hulls to inspect individual rice kernels for damage.
"This is primarily of value for rice breeders and physiologists," says Carl Griffis, professor of biological and agricultural engineering. "They look for traits like resistance to insects and diseases. Current evaluation techniques require removing the hulls for visual inspection of the grain. Those kernels, then, are more difficult to plant to produce the next generation of a breeding line."
Building on the machine vision groundwork of former UA engineer Yang Tao, Griffis and graduate student, Amber Gosnell, are developing a system that can inspect individual rough rice kernels and separate damaged from undamaged. "Yang Tao had an idea to make video images of rice kernels back-lighted with a strong light source," says Griffis. "He demonstrated that damage from disease or insects causes blackened areas on the grains that show up as darkened areas in the images."
A computer program processes those images and predicts damage. Griffis says visual examination of machine-inspected rice shows a high correlation of accuracy. "The machine system is predicting damage with better than 80 percent accuracy. We're working on ways to improve that percentage."
One means of improving accuracy is to add a second camera. Gosnell's research has shown that making two images of each kernel, one each from top and bottom, increases the accuracy.
Other improvements being added to the system include a device that can feed individual kernels onto a transport system that moves them past the cameras one at a time, and a system that uses a puff of air to move damaged and undamaged kernels into separate bins. Improved software will permit "point-and-click" operation of the system.
Karen Moldenhauer, UA rice breeder, has asked Griffis to incorporate a means of measuring the dimensions of kernels still in the hulls. "It's another measure of rice quality that's important to breeders," says Griffis. "Breeders can then plant the undamaged rice to make further selections toward improved varieties.”
Griffis and Gosnell are completing the first year of a three-year study on the machine inspection system. At the end of three years, Griffis plans to deliver a fully automated system to scientists at the Rice Research and Extension Center at Stuttgart.
"We're standing on the shoulders of Dr. Tao to develop this system," says Griffis. "His initial work set the stage for us to continue it toward a full-scale prototype that will become a valuable tool in the effort to develop improved rice varieties for Arkansas producers."